Use real headers and structures to define tables

[skliper]

RTS uses a relative time tag that is 32 bit native endian. Current definition of tables uses a uint16 array which makes the table endian specific.

Could instead define a structure with the real contents (headers and commands), union with an array to make it the right size, then use designated initializers. This would allow the compiler to handle the endianness.

Imported from GSFCCFS-1841

[jphickey]

Just discovered this one, its a real problem. The tables are defined as an array of uint16 values, but read by the FSW as an instance of SC_RtsEntryHeader_t as is done here:

SC/fsw/src/sc_rtsrq.c

Line 104 in b59d70a

SC_ComputeAbsTime(((SC_RtsEntryHeader_t *)SC_OperData.RtsTblAddr[RtsIndex])->TimeTag);

But the "TimeTag" value here is a uint32, not a uint16:

SC/fsw/src/sc_tbldefs.h

Line 59 in b59d70a

typedef uint32 SC_RelTimeTag_t;

Therefore, on a little endian machine, if the TimeTag is not 0, the value is incorrect. The default RTS examples (e.g. rts001, rts002) both use a value of 0 for the TimeTag, so the bug gets hidden - it never attempts to send the other 2 commands in the RTS, that have nonzero values here.

I changed CMD1_TIME from 0 to 1, and sure enough, the SC_ComputeAbsTime is called with a value of 0x1000 (65536), rather than 1, because it was incorrectly cast.

It looks like ATS was updated to split the time tag into a 16-bit MS and LS word (a hack, but will at least get the logical value to come through), but RTS was not updated in the same manner.

[skliper]

@jphickey I've got a PR that was just waiting on something to switch to using the actual structures. Whatever I was waiting on is probably in by now if you want me to move this forward. It doesn't fix how RTS/ATS define time differently... but at least it's endian agnostic.